Face Recognition from Low-Resolution Mobile Image Sequences

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current face recognition systems struggle with low-resolution images, especially from moving devices, leading to inaccurate identification and tracking of individuals, as they require high-resolution images to function effectively.

Innovation Solution

A method and apparatus that aggregates recognition data from multiple low-resolution images to construct a composite image, enabling accurate identification and tracking by distinguishing individuals based on facial features and other metadata without conventional location prediction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If high-resolution images are used for face recognition, then identification accuracy is improved, but the requirement for imaging device capability and image quality increases

Engineering Contradiction:
Improveidentification accuracyVSAvoidimage resolution requirement
Core Design Contradiction:
Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent merges multiple low-resolution images into a single composite high-resolution image through aggregation algorithms. By combining facial feature data from multiple frames, the system reconstructs a detailed composite image that maintains identification accuracy without requiring any single image to have high resolution. This directly resolves the contradiction by transforming multiple low-resolution inputs into a high-quality output.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The system performs preliminary face detection and feature extraction from multiple low-resolution images before generating the composite image. By pre-processing and aggregating recognition data from multiple frames in advance, the system prepares sufficient feature information to achieve accurate identification even when individual images are low-resolution. This preliminary action enables the system to overcome the resolution limitation.

Inventive Principle:
Principle #10Preliminary action

2Adaptability or versatility

If moving imaging devices are used for surveillance, then coverage and flexibility are improved, but image resolution and tracking accuracy deteriorate

Engineering Contradiction:
Improvedevice mobilityVSAvoidtracking accuracy
Core Design Contradiction:
Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent combines multiple low-resolution images captured by moving devices into a composite high-resolution image. By aggregating facial recognition data from multiple frames captured during device movement, the system maintains tracking accuracy despite the low resolution of individual images. This merging process compensates for the degradation caused by device mobility.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The system continuously captures and processes multiple images in sequence, maintaining continuous tracking of the target. By performing face detection and feature extraction across multiple consecutive frames, the system accumulates sufficient recognition data to maintain accurate tracking even when the imaging device is moving and individual images are low-resolution. This continuous action bridges the gaps caused by device mobility.

Inventive Principle:
Principle #20Continuity of useful action

3Measurement precision

If multiple low-resolution images are aggregated to improve resolution, then identification accuracy is improved, but processing complexity and computational resources increase

Engineering Contradiction:
Improveidentification accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts and aggregates only the essential facial recognition data and features from multiple images, rather than processing every pixel and detail. By focusing on key facial features and recognition data points, the system reduces the amount of data that needs to be processed while still achieving high identification accuracy. This extraction approach lowers processing complexity compared to full image processing.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The system changes the parameter of image resolution by aggregating data from multiple low-resolution images to create a high-resolution composite. This parameter transformation allows the system to achieve high-resolution output without requiring high-resolution input images, thereby reducing the computational burden of capturing and processing high-resolution images directly.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3814989B1Method and apparatus for facilitating identification
Publication Date: 2025.11.19 WILDFACES TECH LTD
  • EP3814989B1 patent drawingFigure 1
  • EP3814989B1 patent drawingFigure 2
  • EP3814989B1 patent drawingFigure 3

AI summary

The present invention relates to a method and apparatus for facilitating identification, particularly where images are of relatively low resolution and also where images may be taken by a moving imaging device, such as a mobile camera. In order to identify persons from images of persons, it is usually necessary to obtain a relatively high-resolution image from which face recognition data can be obtained. The face recognition data can be then matched against a database of many persons face recognition data obtained by an equivalent registration process. Where the images are of relatively low resolution, however, it is difficult to obtain reliable matching. The system and method described here, obtain a number of images of a person to be identified and consolidate data obtained from the images to increase the accuracy of identification. This system can identify persons even where relatively low resolution images and mobile cameras.